Alizadeh, Elaheh, authorPrasad, Ashok, advisorDeLuca, Jennifer, committee memberMunsky, Brian, committee memberSnow, Christopher D., committee member2018-09-102019-09-062018https://hdl.handle.net/10217/191378In spite of significant recent progress in cancer diagnostics and treatment, it is still the second leading cause of death in the United States. Some of the complexity of cancer arises from its heterogeneity. Cancer tumors in each patient are different than other patients. Even different tumors from one patient could differ from each other. Such a high diversity of tumors makes it challenging to correctly characterize cancer and come up with the best treatment plan for each patient. In order to do that, a complex combination of clinical and histopathological data need to be collected. This dissertation provides the evidence that the shape of the cells can be used in conjunction with other methods for a more reliable cancer characterization. In this study, experimental studies, numerical representation of the cell shape, big data analysis methods, and machine learning techniques are combined to provide a tool to better characterize cancer cells using their shape information. It provides evidence that cell shape encodes information about the cell phenotype, and demonstrates that the former can be used to predict the latter. This dissertation proposes detailed quantitative methods for quantifying the shape and structure of a cell and its nucleus. These features are classified into three main categories of textural, spreading and irregularity measures, which are then sub-categorized into nine different shape categories. Textural measures are used to quantify changes in actin organization for the cells perturbed with cytoskeletal drugs. Using the spreading and irregularity measures, it is shown that the changes in actin structure lead to significant changes in irregularity of the boundary of a cell and spreading of the cell and nuclei. Using these methods, the shape of retina, breast, and osteosarcoma cancer cells are quantified and it is shown that the majority of cells have similar changes in their shape once they become cancerous. Then, a neural network is trained on the shape of the cells which leads to an excellent prediction of class of cancer cells. This study shows that even though cancer cells have different characteristics, they can be categorized into clinically relevant subgroups using their shape information alone.born digitaldoctoral dissertationsengCopyright and other restrictions may apply. User is responsible for compliance with all applicable laws. For information about copyright law, please see https://libguides.colostate.edu/copyright.cell shapeshape quantificationcancer metastasistexture quantificationmachine learningText